Internal 32K RAM mod - Help

[Proteus]

I have removed the LS00 and the problem continues. I have exactly this:

pin 1(A14) with ground (pin 14 of IC4)
pin 2 (A12) with A12 , this is the cathode of D3 on the ZX81 pcb
pin 22 with ground
pin 23 (A11) with A11, this is the cathode of D1 on the ZX81 pcb
pin 26 (A13) with A13, this is the cathode of D5 on the ZX81 pcb
if you want to run HRG8, connect pin 27 via a resistor of 10K ohms with 5 volts

[RetroTechie]

I'm out of ideas on this one... seems like either there's a damaged circuit trace / short circuit somewhere, changes weren't done as described, or the SRAM is damaged. Or the SRAM uses some non-standard timing so it's incompatible with other low power SRAMs out there?

Compiled a text with pin use for the 32K SRAM as I think you have it now:

Code: Select all

GND         1      *        28 +5V
A12         2      *        27 /WR + 10K pullup
A1          3 (1)    * (24) 26 A13
A9          4 (2)      (23) 25 A5
A8          5 (3)      (22) 24 A3
A7          6 (4)    * (21) 23 A11
A6          7 (5)    * (20) 22 /OE = GND
A4          8 (6)      (19) 21 (Axx) +5V via LK1 or A10 via LK2
A2          9 (7)      (18) 20 /CE = /RAMCS
A0         10 (8)      (17) 19 D4'
D0'        11 (9)      (16) 18 D3'
D1'        12 (10)     (15) 17 D7'
D5'        13 (11)     (14) 16 D2'
GND        14 (12)     (13) 15 D6'

Pin numbers between brackets are for 24-pin IC / socket, makes it easier to refer to ZX81 schematic. Otherwise ignore. Pins marked '*' should be bent sideways & not touch / be in the socket. Please correct if you think there's anything different. AFAICT, all required address lines are present, each only once (!), and go to an address input on the SRAM. Chip select, /OE (grounded) and /WE wiring looks okay. Can you verify that:

• Configuration as you have it now, exactly matches the above?
• Pin 21 on the socket (SRAM) connects to A10 (Z80 pin 40) via LK2 jumper wire, and LK1 jumper wire is removed? (named L1 / L2 on Issue One boards). Otherwise you'd have A10 hardwired to +5V which would explain a lot.

If that's both correct, try this configuration with and without that 33 pF capacitor on /RAMCS pin.

[Proteus]

Hi again, while checking what RetroTechie suggested and having no success, being the processor the main memory handler i decided to change the Z80. It worked like a charm!! Tomorrow i will try to use the LS00 and to remove the cap suggested by Andy. I want to thank you all who helped me and hope this can help other people.

[Proteus]

Tested today with the LS00 and there where some random pixels, not many but after adding one more 47nF to the 33nF in parallel the graphics are fantastic and clean!
Here some pics of the board (messy, must unsolder and make it pretty ) and a screen shot of a HRG demo.

SNC00035.jpg

(1.22 MiB) Downloaded 4531 times

SNC00034.jpg

(1.79 MiB) Downloaded 4530 times

[Andy Rea]

Good Show, glad to see you got it working, the caps just stretch the OE a tiny bit. todays modern memory chips tri-state there outputs much faster than the old ones.

Anyhow, Stay with us, there's usually something interesting going on

Regaards Andy

[RetroTechie]

Yes, congrats for finally working out the problem(s). I think 1st order of business now would be to load up Bodo Wenzel's MaxDemo, which uses both true Hi-Res (larger than regular ZX81 display) and 24K RAM. Wish there was a plain 16K version of this demo, but perhaps the amount of graphic data made that difficult? Should be an excellent test for this mod & the video output.

For 'cleanup', use some SMD parts (perhaps even the LS00). And perhaps you can reduce that capacitor value a little without re-introducing Hi-Res problems. Like a single capacitor of 47, 56 or 68 pF.

(..) the caps just stretch the OE a tiny bit. todays modern memory chips tri-state there outputs much faster than the old ones.

Very possible. This would imply that the ULA hasn't yet read the pixel data by the time it deactivates /RAMCS signal. I was gonna cry "shoddy ULA design", but then I realized the ZX81 was never designed to get the pixel data from the RAM. Or even be able to produce Hi-Res screens, for that matter.

[TMAOne]

the caps just stretch the OE a tiny bit. todays modern memory chips tri-state there outputs much faster than the old ones.

What is OE? Is timing my problem too?

I am trying to do the 32K mod on a Timex motherboard, (again the one on ZX-TEAM's website), with a HY62256ALP-70 chip. It doesn't seem to work. No boot. I have only gotten as far as configuring it for the base 16K, because testing at that point has proved dud.

I have checked the wiring multiple times. As a test for "external" factors, I lifted the RAM chip out, leaving the wires connected, and set the 2K original back in. Boots ok, so I haven't created any shorts or burned out yet another ULA. (Whew.) I bought two of the RAM "just because", so I tried the spare. No difference. Based on the info in this thread, I tried adding a 33 pF cap, but as I have no memory working at all I didn't really expect miracles.

Looking at datasheets I don't really understand, I see a speed of 70 ns, (which I'm pretty sure is nano-seconds). Looking for information on the orginal Toshiba TMM2016P got me a chart with Japanese or Chinese headings but I get "impressions" of figures more in the 100 to 150 range. I have the feeling I'm trying to get The Flash to go fly fishing and he's not impressed.

Neat but dud

Neat but Dud 32K_2.jpg (63.91 KiB) Viewed 8666 times

I did this mod years ago on a ZX81, but it was a Frankenstein job with a narrow 28-pin RAM in a socket with long legs, which were then bent to fit into the wider socket, with the occasional jumper where I broke something altogether and Scotty had to bypass the dilithium converter assembly,... you get the idea. That board is in a custom case so there's room for my ugly bits. I didn't even have proper hookup wire at the time. Picture attached for your amusement.

Frankenmod

Frankenstein 32K_2.jpg (62.13 KiB) Viewed 8666 times

I found a deal on eBay for a used board with no less than eight D43256C-12L chips soldered in. Does that mean 120ns? Any guidance you guys could give me would be appreciated.

Ian

[Proteus]

Hi, do you have pin 1(A14) to ground (pin 14 of IC4) ?

[sirmorris]

You can get vintage '486 motherboards stuffed with 0.3" 32k cache RAM chips. They're usually socketed too. I wouldn't waste time desoldering the eBay board.

[sirmorris]

You can get vintage '486 motherboards stuffed with 0.3" 32k cache RAM chips. They're usually socketed too. I wouldn't waste time desoldering the eBay board.